Decoding orbital angular momentum in turbid tissue-like scattering medium with deep learning

Abstract

Abstract Structured light beams carrying orbital angular momentum (OAM), such as Laguerre–Gaussian modes, are promising tools for high-capacity optical communications and advanced biomedical imaging. However, multiple scattering in turbid media distorts their phase and amplitude, complicating the retrieval of topological charge. Using experimentally acquired three-channel intensity and interference measurements from 25 independent acquisition sessions, we evaluate signed 11-class and unsigned 6-class topological-charge classification with a matched CNN baseline, an Angular Fourier Transform CNN (AFT-CNN), and a pretrained ResNet18 baseline. The best-performing models achieve high accuracy in the low-scattering regime, with the CNN and ResNet18 remaining near 95% at $$z/l^* = 2,$$ but accuracy drops sharply around $$z/l^* \approx 4.$$ These results indicate that sign-dependent OAM information can survive multiple scattering in the low-scattering regime and can be decoded from three-channel measurements with deep learning.